UK judge rejects Stability AI’s bid to dismiss Getty suit
“Inconsistent, or inaccurate” evidence submitted by AI image tool’s CEO | Getty's amended complaint concerning image-to-image feature gets the green light | Justice Smith criticises ‘moveable feast’ of counter-arguments.
A UK judge has dismissed developer Stability AI’s efforts to fend off a copyright infringement claim brought by Getty Images, highlighting credibility issues with the evidence submitted by the AI company's founder, Mohammad Emad Mostaque.
In a decision delivered on December 1, Justice Joanna Smith said there were “obvious question-marks over the credibility of Mostaque’s evidence—evidence which it is essential the court should accept at face value if the defendant has any hope of success”.
As Mostaque faces calls from investors to step down from his role as CEO, Smith found that other available evidence comprising YouTube interviews and billing statements cast doubt over “the credibility of his statements generally”.
The high-profile dispute concerns Stability AI’s deep learning AI model, Stable Diffusion, which is available via a commercial platform, Dream Studio.
Stable Diffusion generates synthetic image outputs in response to commands entered by users.
Evidence of UK training data ‘unsatisfactory’
Getty accuses Stability AI of “scraping” millions of images from Getty Images, without its consent, and using those images unlawfully as input to train and develop Stable Diffusion.
The image library further alleges that the output of Stable Diffusion in the form of synthetic images is also itself infringing because it reproduces a substantial part of the copyrighted works and/or bears their trademarks.
While Stability AI has not yet filed a defence to the claim, it submitted an application seeking a reverse summary judgment and/or a strike out of the complaint.
In summing up, Justice Joanne Smith pointed out that the question at the heart of the application relates to the training and development of the AI tool—and whether there is any prima facie evidence that the training and development of Stable Diffusion took place in the UK.
If there is no such evidence, Stability AI argues “that the claim is bound to fail”, but Justice Smith found that she “remained unsatisfied” with the arguments allegedly supporting this evidence.
Conflict of evidence
Pointing to Mostaque’s evidence stating that UK-based employees didn’t work on the training and development of Stable Diffusion, she said there are contemporaneous documents which “appear to tell a different story”.
Such documents include YouTube interviews where Mostaque seemed to suggest that contractors working on Stable Diffusion were UK-based alongside evidence suggesting he was “personally involved” in the tool’s creation, “notwithstanding his evidence to the contrary”.
These documents, she added, “raise the spectre that Mr Mostaque’s evidence is either inaccurate, or incomplete”.
Smith determined that “at the very least they suggest a conflict of evidence and thus reasonable grounds for believing that a fuller investigation into the facts of the case would add to or alter the evidence available to the trial judge in respect of the location issue”.
In doing so, she rejected Stability AI’s submission that Mostaque’s evidence “is “clear or definitive”, amid “unexplained evidence of data transfers to London” from the US as well “as significant issues concerning the evidence over when work on Stable Diffusion actually commenced”
Image-to-image claim ‘should go to trial’
Justice Smith then turned to Getty’s amendment application seeking to introduce a new case concerning an image-to-image feature enabling Stable Diffusion to generate a synthetic image output in response to an image uploaded by a user—either with or without a text prompt.
The proposed amendment introduces claims of copyright infringement, database right infringement, trademark infringement and passing off, in relation to the image-to-image feature.
Describing Stability AI’s objections to the amendments as “something of a moveable feast”, Smith noted that the defendant’s key counterargument appears now to be that when a copy of Stable Diffusion is downloaded by a user from GitHub with a view to then entering the image prompt, it is in fact the user who brings about the copying.
Such an action, according to Stability AI, cannot possibly amount to infringement on its part.
However, Justice Smith concluded that the image-to-image claim as pleaded “has a real prospect of success” and that “it is plainly a claim that should go to trial”.
She further found that Stability AI had failed to explain its counterarguments in good time, and had not explained them fully.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox
